Investigate inconsistency between sncosmo and our band fluxes

[hombit]

Updated in PR 132 SN Ia end2end test and notebooks show small but significant few percent inconsistency between our band fluxes and their. The problem is especially bad when negative flux densities are presented.

I'm not sure what is the root or the problem, some possible options:

• Difference between trapezoid integration we use and rectangle integration sncosmo uses
• Difference in interpolation
• Some problems in our band integration code
• Some problems in our redshift code